Context-enriched interactome powered by proteomics helps the identification of novel regulators of macrophage activation
- Channing Division of Network Medicine, Brigham and Women’s Hospital, Harvard Medical School, United States
- Center for Interdisciplinary Cardiovascular Sciences, Brigham and Women’s Hospital, Harvard Medical School, United States
Figures
Figure 1 with 1 supplement
Co-abundance networks.
(A) Top row: Sample from the Pearson correlation matrix showing the top 40 proteins in alphabetical order. Pearson’s r values are calculated based on the protein abundance profiles for each …
-
Figure 1—source data 1
Source data for Figure 1A: Co-abundance networks in edgelist format (Columns: Protein A - Protein B).
Co-abundance network for M(-).
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig1-data1-v2.csv
-
Figure 1—source data 2
Source data for Figure 1A: Co-abundance networks in edgelist format (Columns: Protein A - Protein B).
Co-abundance network for M(IFNg).
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig1-data2-v2.csv
-
Figure 1—source data 3
Source data for Figure 1A: Co-abundance networks in edgelist format (Columns: Protein A - Protein B).
Co-abundance network for M(IL-4).
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig1-data3-v2.csv
-
Figure 1—source data 4
Source data for Figure 1C: Odds ratios and confidence intervals, and p-values for the link validation sets shown in Figure 1C.
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig1-data4-v2.xlsx
Figure 1—figure supplement 1
(A) Pearson’s r correlation threshold used in filtering co-abundance edges, as a function of the edge density of the resulting co-abundance network, at FDR = 0.01. (B) The degree distributions P(k) …
Figure 2 with 10 supplements
Target candidate prioritization and its performance assessment.
(A) The entire literature curated PPI network with co-abundance edges from all three stimulation conditions, providing a global view of the distribution and connectivity of co-abundance edges and …
Figure 2—figure supplement 1
(A) Significance of the largest connected component (LCC) of CVD drug targets mapped onto the PPI network only and the respective combined networks. Z-scores are calculated based on 10,000 …
Figure 2—figure supplement 2
For the iCTNet CVD drug targets.
(A) The k-fold cross-validated (k = 7) ROC curves for M(-), M(IFNγ), M(IL-4), PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), respectively. Shaded regions represent the standard deviation of the …
Figure 2—figure supplement 3
For the iCTNet CVD drug targets.
(A) ROC curves for M(-), M(IFNγ), M(IL-4), PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), comparing the original PPI, PPI with low-throughput edges removed, and PPI with low-throughput and …
Figure 2—figure supplement 4
(A) The weight ratio between co-abundance and PPI edges, w(CoA)/w(PPI), as a function of the area under the ROC curve (AUROC) for PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), respectively. Shaded …
Figure 2—figure supplement 5
(A-C) For the DrugCentral CVD drug targets. (A) The k-fold cross-validated (k = 7) ROC curves for M(-), M(IFNγ), M(IL-4), PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), respectively.Shaded …
Figure 2—figure supplement 6
(A-C) For DrugCentral all drug targets. (A) ROC curves for, PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), respectively.(B) Violin plots showing the distribution of percentile ranks of drug …
Figure 2—figure supplement 7
(A-C) For the DrugCentral all drug targets. (A) The k-fold cross-validated (k = 7) ROC curves for M(-), M(IFNγ), M(IL-4), PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), respectively.Shaded …
Figure 2—figure supplement 8
(A) ROC curves for PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4) where the co-abundance networks M(-), M(IFNγ) and M(IL-4) were built using biweight midcorrelation as the correlation measure. (B) …
Figure 2—figure supplement 9
(A) ROC curves for PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4) for the validations sets, where random walk with restarts (RWR) was used as the prioritization method. (B) Venn diagrams showing …
Figure 2—figure supplement 10
(A-C) For the inflammation (Inflammatome) signatures. (A) The k-fold cross-validated (k = 7) ROC curves for M(-), M(IFNγ), M(IL-4), PPI, PPI + M(-), PPI + M(IFNγ) and PPI + M(IL-4), …
Figure 3 with 3 supplements
Pathways of top prioritized proteins.
(A) The Venn diagram showing the shared pathways between the significantly enriched (q-value <0.05) pathways of the top candidates (empirical p-value<0.01) prioritized by the addition of the M(-), …
-
Figure 3—source data 1
Source data for Figure 3: All enriched pathways of the top N (p<0.01) prioritized proteins.
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig3-data1-v2.xlsx
Figure 3—figure supplement 1
The full list of hierarchically clustered condition-specific pathways.
Darker shades of colors indicate a higher enrichment in terms of the negative logarithm of q-value.
-
Figure 3—figure supplement 1—source data 1
The enrichment p-values of the condition-specific pathways shown in the clustered heatmap of Figure 3—figure supplement 1.
- https://cdn.elifesciences.org/articles/37059/elife-37059-fig3-figsupp1-data1-v2.xlsx
Figure 3—figure supplement 2
The network of significantly enriched pathways of the top-ranked (empirical p-value<0.01) candidates prioritized by the addition of the M(IFNγ) co-abundance network to the PPI network.
Node size indicates the significance of each pathway in –log(q-value) and edge thickness represents the overlap between the genes of the two pathways they connect, measured in terms of the Jaccard …
Figure 3—figure supplement 3
The network of significantly enriched pathways of the top-ranked (empirical p-value<0.01) candidates prioritized by the addition of the M(IL-4) co-abundance network to the PPI network.
Node size indicates the significance of each pathway in –log(q-value) and edge thickness represents the overlap between the genes of the two pathways they connect, measured in terms of the Jaccard …
Figure 4 with 3 supplements
Selection of top target candidates by incorporating expression and abundance data.
(A) Overview of the filtering/combined ranking scheme. Three rankings were performed according to: (i) the network prioritization, (ii) the relative protein abundance difference with respect to …
Figure 4—figure supplement 1
(A) Optimal threshold on the network-based prioritization ranking maximizing the sum of sensitivity and specificity. The optimal threshold value was the top 2971 candidates for M(IFNγ).(B) Optimal …
Figure 4—figure supplement 2
(A) The Spearman’s correlation of combined ranking with each constituent ranking, that is gene expression, network prioritization, and protein abundance, respectively, for M(IFNγ). (B) The …
Figure 4—figure supplement 3
(A) The Spearman’s correlation between pairs of constituent rankings for M(IFNγ). (B) The Spearman’s correlation between pairs of constituent rankings for M(IL-4).
Figure 5 with 3 supplements
In vitro loss-of-function experiments for GBP1 and WARS.
(A) Relative mRNA and protein expression levels for CCL2, TNFα, JAK2 and STAT1 and with WARS silencing under control and IFNγ stimulation conditions, performed on THP-1-differentiated …
Figure 5—figure supplement 1
(A) Protein expression level of CCL2 with WARS silencing under control and IFNγ stimulation conditions, performed on human primary PBMC-derived macrophages. *p<0.05, Student’s t-test. (B) Western …
Figure 5—figure supplement 2
(A) The number of total, tractable, druggable and ensemble druggable structures in the top candidates, given by the DrugEBIlity database. (B) The percentage of tractable, druggable and ensemble …
Figure 5—figure supplement 3
(A) Spearman correlation values between the gene expression and protein abundance for the top 2826 candidates for M(IFNγ) and the top 4109 candidates for M(IL-4), based on maximizing the sum of …
Additional files
-
Supplementary file 1
Topological properties of the PPI network and co-abundance networks.
- https://cdn.elifesciences.org/articles/37059/elife-37059-supp1-v2.docx
-
Supplementary file 2
Overlap of co-abundance networks with large-scale binary and affinity-purification-mass spectrometry (AP-MS) based interactomes (Fisher's exact test, two-sided p-values).
- https://cdn.elifesciences.org/articles/37059/elife-37059-supp2-v2.docx
-
Supplementary file 3
Overlap of the top prioritized (p<0.01) proteins with inflammation, innate immune response, and coronary artery disease signatures (Fisher's exact test, two-sided p-values).
- https://cdn.elifesciences.org/articles/37059/elife-37059-supp3-v2.docx
-
Transparent reporting form
- https://cdn.elifesciences.org/articles/37059/elife-37059-transrepform-v2.docx
